Andersons pediatric cardiology 636

■ Is there evidence for a dose-response relationship? ■ Is the relationship specific for the risk factor and outcome of interest? ■ Is the relationship consistent across variations in study populations, settings and investigators, and design? ■ Is the relationship free of known and potential confounding? ■ Is the relationship free of systematic and random measurement error? ■ If an intervention is successful in reducing or eliminating the risk factor, does this alter the outcome in a consistent and predicted manner? Experimental Studies and Trials Randomized Clinical Trials Randomized controlled trials remain the gold standard for evaluating the efficacy and safety of treatments or interventions (see Table 24.2) The principle of a randomized controlled trial is simple A group of subjects is randomly assigned to an experimental arm (or arms) or to a comparison arm The comparison arm can be placebo or an active control (such as the standard-of-care medication for the condition being treated) Randomization markedly reduces the risk of systematic differences in characteristics at baseline between the groups Any outcomes observed during the period of follow-up, therefore, should be attributable to the treatment alone and not to underlying differences in subjects Randomized clinical trials provide the highest grade of evidence, particularly regarding establishing a causal relationship between an intervention and an outcome This is true for two primary reasons First, randomization provides the best chance that the groups chosen for study will be equivalent at baseline regarding both measured and unmeasured characteristics Any differences would be attributed to chance or random error The larger the population studied, the less likely random maldistributions are to occur Strategies can be used to ensure equal numbers of subjects in groups, so-called block randomization, and to ensure that key characteristics are equivalent in groups, so-called stratified randomization When assessing the success of randomization by comparing measured characteristics at baseline, it is important to note clinically meaningful differences between groups rather than statistically significant differences If the sample is large, small and unimportant differences may be statistically significant If the sample is small, large and important differences may not be statistically significant The second feature of randomized clinical trials that contributes to providing high-quality evidence is that all clinical trials are based on prospectively and concurrently collected data Thus there is opportunity for standardization of definitions, methodology, measurements, reporting, and implementation of measures for the control of quality, such as adjudication An important aspect of assessment is the concept of blinding or masking, whereby the subjects, the investigators, other study personnel, and those analyzing the data have no knowledge of the randomized assignment of the subjects This ensures that the protocol is implemented identically in all groups Additionally, even the most well-meaning assessors may interpret subjective tests differently if they think that they know the assignment of the subject It may not be feasible for some interventions to be blinded completely For instance, for a trial of surgical versus transcatheter closure of septal defects, it would be obvious to patients and providers which treatment a patient received In this case, however, some blinding could still be used, as measurements, assessments, and analysis of the data can almost always remain blinded There are multiple phases in clinical trials depending on the amount of previous knowledge available regarding a specific intervention Also, there are different types of clinical trial approaches to be used depending on the aim to be achieved Clinical trials can be divided into four phases based on their aims Phase I trials establishes the safety of the intervention, phase II trials establish therapeutic effects, phase III trials are a full-scale evaluations of an intervention by comparing the intervention to a control and establishing relative benefit and risks, and phase IV trials evaluate the long-term effects of an intervention, often called postmarketing trials.46,47 Prior to performing a clinical trial, it is important to determine which phase of the intervention the trial will represent and what evidence will be needed The first and second phases are often small trials, using fewer than 50 patients, and aimed at establishing safety and therapeutic effect For drug trials, they often also incorporate assessments of pharmacokinetics and pharmacodynamics They rarely contribute to major changes in clinical practice The third and fourth phases strive to compare the intervention with a standardof-care practice, a placebo, or a nonintervention These trials are the driving force behind changes in clinical practice and important components of the evidence base supporting guidelines Since the methodology for designing, performing, analyzing, and reporting clinical trials is fairly standardized, this makes them easier to appraise In the past 10 years, many biomedical journals have adopted the CONSORT guidelines for the standardized reporting of clinical trials in the scientific literature.48 Meta-Analyses Meta-analyses and systematic reviews, in an effort to raise the certainty of a finding, pool data from multiple clinical studies to try to establish whether multiple studies of a topic actually have similar findings (see Table 24.2) Metaanalyses identify a problem of interest and compile all studies, whether observational or a randomized clinical trial, whether producing positive or negative results, and whether published or not, that have previously studied the issue at hand Results from each study are then critically appraised and compiled and a summary conclusion is achieved.49,50 Meta-analyses are used to decrease the effect of studies with limited validities that might have inconsistent results Statistical analysis of meta-analyses is highly complex and beyond the scope of this chapter Peer-reviewed meta-analyses are published in The Cochrane Review, a repository of meta-analyses and systematic reviews, which is a highquality reference on which to base changes in clinical practices